February 25, 2019

Good Morning,

I've always been leery about machines, especially when they have the ability to learn. And researchers are backing me up and still instisting that humans double check robot work. Never stop checking.

Learn about this and more interesting stories from the scientific community in today's issue.

Until Next Time,
Erin

P.S. Did you miss an issue? You can read every issue from the Gophercentral library of newsletters on our exhaustive archives page. Thousands of issues, all of your favorite publications in chronological order. You can read AND comment. Just click GopherArchives

*-- Machine learning-based discoveries still need to be checked by humans --*

Researchers at Rice University want scientists to continue double-checking discoveries made using machine learning.

Until machine-learning systems are capable of self-critique, scientists warn their predictions can't be fully trusted.

"Work is underway on next-generation machine-learning systems that will assess the uncertainty and reproducibility of their predictions," Rice University statistician Genevera Allen said in a news release.

Machine-learning systems are designed to make predictions about future data given what they've learned by analyzing current datasets.

"A lot of these techniques are designed to always make a prediction," she said. "They never come back with 'I don't know,' or 'I didn't discover anything,' because they aren't made to."

Machine-learning systems are currently used to develop cancer drugs targeting patients with similar genomic profiles.

"People have applied machine learning to genomic data from clinical cohorts to find groups, or clusters, of patients with similar genomic profiles," Allen said.

But often, these studies produce uncorroborated results -- findings that aren't easily replicated.

"The clusters discovered in one study are completely different than the clusters found in another," Allen said. "Why? Because most machine-learning techniques today always say, 'I found a group.' Sometimes, it would be far more useful if they said, 'I think some of these are really grouped together, but I'm uncertain about these others.'"

Allen shared her analysis of machine-learning systems and their biases at the 2019 Annual Meeting of the American Association for the Advancement of Science, held over the weekend in Washington, D.C.

*-- Worms help scientists understand memory formation and recall --*

The 302 nerve cells inside roundworms are helping scientists understand why some events or associations can't be remembered.

Memory blocking, or Kamin blocking, describes an animal's inability to replace an old memory cue with a new one if the new cue is presented at the same time as the old one.

"Suppose you grew up hearing ice cream trucks playing a song and hearing that song, even when you can't see the truck, makes you think of ice cream," Daniel Merritt, professor of molecular genetics at the University of Toronto, said in a news release. "One day, the ice cream truck owners decide to add a spinning green light to the roof of the truck, so that even people who are hard of hearing can see them. Kamin blocking predicts that you won't learn to associate spinning green lights with ice cream, because the ice cream truck song already fully predicts the delicious treat in store for you."

Psychologists have pointed to the Kamin blocking as evidence of the importance of novelty for the process of memory formation in humans.

While evidence of Kamin blocking can be observed externally among mammals, understanding its neurological underpinnings has proven especially difficult. Worms offer a simpler test model.

"Being able to fully describe the molecular changes that are going on in memory is enormously appealing, but human memory is too ephemeral and nebulous to pin it down," said Merritt. "But by studying it in worms, we are really making a lot of headway in figuring out exactly what is going on when memories are formed and retrieved in a molecule by molecule fashion."

Before Merritt and his colleagues could study Kamin blocking in worms, they had to confirm the simple roundworm experiences memory blocking.

Researchers trained worms to associate hunger with either the taste of salt or the aroma of benzaldehyde, a compound with an almond-like odor. Roundworms typically find the taste of salt and smell of benzaldehyde appealing, but trained worms are repelled by the cues.

Next, scientists exposed benzaldehyde-trained worms to both benzaldehyde and salt. When the worms were exposed to only salt, they continued to crawl toward the taste. Salt-trained worms continued to find benzaldehyde appealing, even after being exposed to salt and benzaldehyde together.

The first cue prevented the formation of a second association.

Next, scientists used green fluorescent molecules to tag and highlight the expression of the EGL-4 protein expressed by a single olfactory nerve cell in the heads of the test subjects. EGL-4 is essential for benzaldehyde starvation learning.

When researchers repeated the experiments, they found the EGL-4 protein behaved the same way during benzaldehyde blocking as it did during normal learning.

"This is interesting because it contradicts the classic interpretation of blocking where you need an element of surprise or you don't bother remembering the second association," said Merritt. "Our data shows that the memory is formed but it's the expression of behavior that's suppressed somehow."

Scientists shared the latest findings in the journal Scientific Reports.

Blocking's effects last four hours, preventing a new association from being learned. Scientists don't know why blocking produces a hangover effect.

"That's the big question," said Merritt. "I'll let you know in four years' time when I'm done with my PhD."